OUP user menu

Why do animal models of post-angioplasty restenosis sometimes poorly predict the outcome of clinical trials?

Antoine Lafont, David Faxon
DOI: http://dx.doi.org/10.1016/S0008-6363(98)00109-6 50-59 First published online: 1 July 1998

Time for primary review 22 days.

The 20th anniversary of percutaneous transluminal coronary angioplasty (PTCA), first introduced by Andreas Gruentzig MD, was recently celebrated. Since its introduction, millions of people have been successfully treated by PTCA. Despite its overwhelming success, 30 to 50% of patients develop restenosis, a rate that has changed little since the introduction of the technique. After 20 years of dramatic technical refinement and intensive research, we still do not know why a dilated artery will maintain patency and not develop restenosis or will develop restenosis. The lack of understanding about restenosis has often been attributed to inappropriate experimental models, incomplete or incorrect analysis of the models that has led to a focus on the wrong pathophysiologic target [1–4]. In the review, we will describe the various animal models used to study restenosis, and clarify the limitations and advantages of each in order to better delineate the ideal model of experimental angioplasty and restenosis. While outstanding reviews have already been written on this subject within the last five years [5–8], this review will be different from the former ones because it will take into account the following major changes: Stent-related restenosis, which is different from balloon-related restenosis, and arterial remodeling [9–12]. Arterial remodeling has transformed our understanding of restenosis and is of great clinical relevance, as shown by the good concordance between experimental data and human data obtained by intravascular ultrasound [13–19].

1 Experimental models

Before angioplasty started, balloon abrasion with endothelial denudation represented the gold standard for studying the response to vascular injury and smooth muscle cell proliferation and migration [20–25]. The idea of balloon angioplasty seemed to be a questionable method, and the occurrence of restenosis was not a surprise. Thus, experimental models of restenosis preexisted the clinical recognition of restenosis after angioplasty.

All experimental models of restenosis require a living animal in which an artery has been injured by a balloon catheter or some other means. The differences between each model concern mainly the animal species, the size of the artery lumen, the presence or absence of atherosclerosis, the extent of thrombogenicity of the model and the degree of the balloon injury. The most commonly used animal models include the rat, rabbit and swine.

2 The rat model

This is the first model used since it was developed to study the response to injury and atherosclerosis and, therefore, preceded the development of angioplasty. The lesion is created by abrasion of the internal carotid artery by a Fogarty balloon being surgically introduced via the external carotid artery [22, 26]. The resulting lesion is usually evaluated by histology two weeks after angioplasty and is characterized by a pure neointimal hyperplasia with smooth muscle proliferation and matrix deposition (Fig. 1). A lighter injury, limited to the endothelium, by a nylon filament results in a mild proliferative response [20, 27]. The rat model represents the easiest model to use, because it is very cheap, lesions develop rapidly, and it requires minimal equipment and does not require a catheterization laboratory. It was used initially as an atherosclerotic model to study smooth muscle cell proliferation and migration and extensive literature exists concerning these events [21, 23, 28–35]. In the setting of angioplasty, the model was used to test the efficacy of strategies targeting smooth muscle cell proliferation [35–41]. The disadvantages of this model are the following: It utilizes a normal artery, and the balloon injury does not reflect the angioplasty procedure, since the latex Fogarty balloon significantly overstretches a normal artery at low pressure, whereas the PVC balloon angioplasty catheter maintains a balloon-to-artery ratio of 1.2:1. The histology of the lesion is unlike human atherosclerosis as it lacks several features of the human atherosclerotic lesion, such as calcification and calcium deposits, and it is unlike human restenosis since it occurs in a normal artery.

Fig. 1

Photomicrograph of a rat carotid artery 14 days after balloon injury. There is a discrepancy between the size of the vessel and the lumen.

3 The rabbit model

This model, unlike the rat model, is usually a double injury model. The first injury serves to develop an atherosclerotic lesion that will undergo a second injury i.e., the angioplasty. The first lesion often, but not always, is developed by use of a high cholesterol diet (classically 6% peanut oil and 2% cholesterol). The initial injury is usually developed in two possible ways, either balloon abrasion or air-desiccation [42–45]. The advantage of air-desiccation is that it clearly provides a well defined focal lesion site [13, 42]. Three to six weeks after lesion induction, an angioplasty is performed under angiographic control. Four weeks after angioplasty, an angiogram is performed immediately before sacrifice and histologic evaluation. The features of the lesion are characterized by smooth muscle proliferation, extracellular matrix accumulation and inflammation with an abundant presence of foam cells (Fig. 2). The acute histological findings after angioplasty are characterized by dissection of the artery as well as the presence of thrombus. Thus, the rabbit model mimics relatively well the angioplasty process occurring in humans, and it can be used to study new experimental procedure-, device- or drugs-related strategies [43, 46–68]. It also can be used to evaluate mechanisms of restenosis, extracellular matrix formation and inflammation occurring after angioplasty [13–19, 69–73]. While the model is closer to the human situation, caution must be taken because of significant differences from human restenosis. The major criticism has been the abundant presence of foam cells, although they are also present to a lesser degree in humans. Finally, a less frequently used rabbit model is the ear artery injury model: This model offers the main advantage of being very easy to perform [74]. Ear artery has been used for arterial gene transfer [75]. Typically, it consists of a lesion induced either by air-desiccation or crushing the artery, which is performed without any surgical or x-ray means (Fig. 3). The lesion is mainly based on neointimal hyperplasia [74]. The artery is of easy access. However, the size of the artery lumen (1 mm) renders the model less accessible for angioplasty.

Fig. 3

Photomicrograph of an atherosclerotic ear artery from a rabbit. Adenovirus-mediated lacZ gene transfer β-galactosidase expression is seen in endothelial cells (blue color), but not in the neointima (N), performed four weeks after lesion induction.

Fig. 2

Photomicrograph of a rabbit restenotic iliac artery eight weeks after arterial denudation and high cholesterol diet and four weeks after balloon angioplasty. Magnification ×20. L=lumen; N=neointima.

4 The pig model

Unlike the rabbit model, this model usually is a single injury model [8]. The injury is performed either in the carotid artery or in the coronary artery [76, 77]. Most commonly, the lesion is induced by overstretching the artery via a balloon catheter, which is sometimes pulled back to increase the injury. The balloon artery ratio classically is greater than 1.4. This degree of stretch injury is necessary in order to fracture the external elastic lamina in order to induce a sufficient degree of neointimal hyperplasia (Fig. 4). The degree of tearing of the external elastic membrane correlates with the degree of intimal hyperplasia [78]. An injury score has been proposed, which correlates with neointimal hyperplasia and allows for more reliable comparisons between animal groups [78]. It is important to note that the degree of stretch injury that is used in these models does not reflect what is used in angioplasty procedures in humans [79]. In some cases, the lesion is induced by stent placement in the carotid or the coronary artery. The stent is also overexpanded, which results in a severe stenosis during follow-up. In contrast with rabbits, the pigs are usually not placed on a high cholesterol diet, due to cost and the time required to induce lesion formation [80]. Although the model has been described [82–85], the lesion that forms following the overstretch coronary balloon angioplasty is mainly a neointima with smooth muscle cell proliferation without foam cells [77–86]. There is usually partial or occlusive thrombus formation. The pig model is characterized by a higher thrombogenicity than the other models [87–89]. This can be explained by a lower circulating plasminogen level [89]. Like the rat and the rabbit models, the pig has been mainly used for evaluation of strategies that inhibit smooth muscle proliferation [8, 76, 89–100]. Unlike all other models, the pig model utilizes the coronary vessels, thus allowing optimal evaluation of the angioplasty procedure and the evaluation of new coronary devices [97–100].

Fig. 4

Photomicrograph of a pig coronary artery. The lesion was induced by the use of balloon thermal injury (80°C for 30 s). Reproduced by courtesy of Dr Gregoire et al. from “Arterial remodeling a critical factor for restenosis”. Lafont and Topol, editors, Dordrecht: Kluwer, 1997, p 172.

5 The dog model

As carnivores, dogs do not easily develop atherosclerotic lesions. The reaction to acute injury (angioplasty) or chronic injury (stent) does result in a mild neointimal reaction. However, plasminogen levels are higher than in humans [89]. This model has been used mainly to test interventional cardiology devices, such as lasers and stents, because coronary artery anatomy is similar to human coronary anatomy [101–103]. Another dog model involves the placement of a ligature around the LAD, to induce cyclic flow variations, followed by balloon injury distal to the ligature [104]. Restenotic-like lesions have been shown to develop. However, given the difficulty in creating lesions, this model is infrequently used.

6 The nonhuman primate model

The nonhuman primates (chimpanzee, baboon, rhesus macacus, cynomologus monkey) most closely mimic the human given their close relationship to the human species. In order to induce atherosclerosis, animals are usually placed on a high cholesterol diet [25]. Lesion development can be assessed by noninvasive techniques such as MRI or IVUS. Lesion development usually takes many months or years [105, 106]. The histology of the lesions is the closest to human atherosclerotic lesions (Fig. 5) [107]. However, the cost, the long period of lesion induction, and the difficulty in handling these species makes this model unrealistic for standard use in the evaluation of mechanisms or preventive strategies of restenosis.

Fig. 5

Photomicrograph of a human atherosclerotic coronary artery.

6.1 Interpretation of experimental studies

Restenosis after angioplasty has been classically attributed to neointimal formation due to smooth muscle cell proliferation and matrix secretion [4, 5, 7, 108]. Experimental models were designed to produce neointimal hyperplasia in order to test treatments against smooth muscle proliferation [3, 42, 77]. However, there has been a major discrepancy between the overall success of many interventions in these models and their failure in clinical trials [109–124]. Animal models have been considered responsible for this failure. We will analyze the causes of the apparent failure of these models to lead to a treatment for the prevention of restenosis in humans.

7 Are the models wrong?

It is important to question the validity of the animal models. Are these models not appropriate? Differences in species represent an issue that must not be underestimated. The rat model is not considered to be representative of the angioplasty procedure, and rat and human species are very different. On the other hand, the nonhuman primate appears to be nearly ideal, although it does not involve the coronary vessels. The size of the vessel is also an important issue. The normal rat carotid artery is much smaller in size and is structurally different from the human coronary artery. As Muller et al. [5]pointed out, the response to injury will be different in a muscular or elastic artery. With regard to the size and structure, the rabbit iliac artery comes closer to the human coronary artery. There has also been a misunderstanding between the severe stretching injury induced by angioplasty on the artery wall, and the mild denudation injury necessary for lesion induction to obtain an atherosclerotic lesion. The lack of an atherosclerotic lesion prior to angioplasty is also an important issue with respect to the similarity to man, in which normal arteries are usually not subjected to angioplasty. While it is important to recognize these limitations and differences between the animal models and the human situation, the models can appropriately provide answers to specific questions. For instance, if it is important to determine the mechanism of a drug on smooth muscle cell proliferation, then a number of animal models will be capable of answering this question [51, 77, 86, 125]. The results however should not be interpreted to mean that the drug is also able to inhibit restenosis in man. The extrapolation of animal studies directly to man is unreasonable given the vast differences between animal models and man and the complexity of the restenotic process.

8 Were the analysis or the endpoints wrong?

The methods used in analyzing data from the restenotic models is critical in order to evaluate first the appropriateness of the model and, second, the validity of the data. In order to determine if restenosis has occurred, it is customary to compare the lumen of the dilated area to a reference lumen. While this is the clinical method for defining restenosis, it is surprising to see (in numerous studies) that it is often not reported. The second parameter that is used to define the degree of restenosis is the area of the neointima. While this is frequently reported instead of changes in lumen diameter, it is not a good reflection of restenosis. The explanation for this is the importance given to neointimal hyperplasia as the major mechanism for restenosis. Neointimal thickness is frequently measured to evaluate neointimal hyperplasia, although it represents both cellular content and matrix deposition. It also can be affected by tissue fixation. The third parameter that is measured to assess restenosis is the area circumscribed by the external elastic lamina. This parameter evaluates the ability of the artery to undergo positive or negative remodeling. Since the importance of remodeling was not appreciated prior to 1993, much of the experimental data evaluating various treatments to prevent restenosis failed to take remodeling into account.

9 Discrepancy between animal studies and clinical trials

The discrepancy between positive results in experimental models and overall failure to reproduce these results in humans has been partly explained by the following points: Inadequate dose of drugs, inadequate statistical power, poorly defined endpoints and a focus on less important therapeutic targets.

Currier and Faxon [1]have pointed out the worrisome discrepancy between the drug dosages used in animals and in humans, with the ratio between animal and human studies being two- to 40 times greater. The time when the drug was delivered is also of importance, having a substantially greater effect if pretreatment is given.

Too small a number of patients have been included in most of the clinical trials; this might also be the reason for the lack of statistical significance. This methodological issue was already raised by Muller et al. [5]five years ago and continues to be a significant problem.

Another methodological issue is the difference of endpoints: In animal models, endpoints are mainly the neointimal hyperplasia and lumen areas, whereas angiographic lumen loss or target vessel revascularization have been the clinical endpoints. Recently, with the emergence of intracoronary ultrasound, changes in neointimal thickness and remodeling are now measurable in man as well [18].

10 Misunderstanding of the targets

On the basis of animal studies and human pathological studies, the main hypothesis for the mechanism of restenosis was attributed to intimal hyperplasia. Smooth muscle proliferation was considered to play a major role in restenosis because it was established to occur in experimental models as well as in humans. A link was made between restenosis and neointimal growth, despite the lack of a proven correlation between the degree of neointimal thickening and restenosis. Neointimal development was considered to be a tumor reducing the lumen because the artery was considered to be a “rigid tube” that could not expand or shrink, independently of the arterial growth. It is clear that neointimal growth occurring in a “rigid tube” will irrevocably turn into lumen narrowing. Waller et al. [126]detected the absence of intimal hyperplasia in a number of patients who developed restenosis clinically, from a small series of twenty patients who came to necropsy. The lack of evaluation of the area circumscribed by the external elastic lamina, a measure of remodeling, led to a focus on intimal hyperplasia and smooth muscle proliferation, the wrong target, and helps to explain why experimental models were apparently unable to predict the ability of drugs to inhibit restenosis.

10.1 Arterial remodeling after balloon angioplasty

When it became apparent that arteries could undergo remodeling, measurement of the histomorphometric parameters of the external or internal elastic lamina were used. In all of the models except the rat model, restenosis was found not to be primarily related to neointimal hyperplasia, but to constrictive remodeling and/or a lack of enlargement [12–17, 107]. The greatest advance in our understanding of restenosis has been the understanding of the role of remodeling and the potential new targets to prevent this problem. Conversely, the rigid tube model applies perfectly to in-stent restenosis [127]. The irony has been to focus on neointimal hyperplasia during the era of balloon angioplasty and to discover the role of the remodeling in restenosis at the time of the stent era, since stenting reduces restenosis by inhibiting constrictive remodeling, as shown by the STRESS/BENESTENT trials [10, 11]. While stenting can prevent unfavorable remodeling, it leads to excessive intimal hyperplasia, as shown in the porcine model. The stented vessel in which neointimal hyperplasia occurs inside the stent will behave as a rigid tube, as originally conceived. There is no possibility of enlargement. In a sense, this simplifies restenosis by limiting the process only to intimal hyperplasia and thrombosis. An important consequence of in-stent restenosis is the need for applicable models for this new cause of restenosis, distinct from balloon-related restenosis. The swine stent model proposed by Schwartz et al. [77]five years ago is currently the best model for studying stent restenosis; however, it is probably not ideal. The atherosclerotic rabbit model could offer an interesting complement.

10.2 Perspectives

We have learned a lot from our experimental models about restenosis. We have identified two major forms of restenosis that appear to be independent; balloon related- and in-stent restenosis. Balloon-related restenosis is largely due to unfavorable arterial remodeling. In the future, we need to define new targets to prevent constrictive remodeling and promote positive remodeling. Preliminary work has already been done. Antioxidants have been shown to reduce restenosis in various models, such as the atherosclerotic rabbit and the porcine models [52, 62, 96]. Experimental data suggest that their prevention of restenosis is due to an effect on remodeling. This finding led to the success in reducing restenosis in humans by probucol. In the Multivitamin Probucol Trial, late loss was reduced by 50% and this favorable action was demonstrated to be related to favorable remodeling and to be independent of neointimal hyperplasia, via intracoronary ultrasound [128, 129]. These results should encourage further research into new targets with the help of experimental models. Stent-related restenosis is paradoxically well characterized, but solutions to prevent or reduce it are not yet established. The current focus is more towards local drug delivery than the new treatment options. Development of gene therapy, local irradiation or coated stents might help to overcome the double problem of a healing process and a foreign body [33, 34, 130–134]. Bauters et al.[135]have shown that the human homozygote phenotype DD was clearly related to a higher risk of in-stent restenosis. Further work with transgenic models might also help to detect populations at risk of an intense reaction of smooth muscle cell proliferation [125, 136].


View Abstract